Time Step Error#
Time step error = <float> <integer_list>
Description / Usage#
The time step error controls the adjustable time step size based on the difference between the solution and the predicted solution (L2 norm). The first of the eight arguments is a floating point number that indicates the error in the time step selection.
- <float>
the error value, any floating point number.
The smaller this number is, the smaller the time step will tend to be in the automatic time step control. The original implementation of this capability in Goma did not use a normalized value for the norm; to enable this most useful feature, use a negative value of the time step error and a positive, normalized norm will be computed. This way a percentage value of the solution error will be set.
- <integer_list>
seven integers, with a value either zero (0) or one (1).
A further degree of control is offered by the seven integers (i1 through i7) that identify which solution variables will contribute to the error norm calculations. Permissible values for each of these seven integers are 0 and 1. The correspondence between the integers and variables is as follows:
i1
(pseudo) solid displacement
i2
fluid velocity
i3
temperature
i4
concentration, porous liquid pressure, gas pressure, porosity, saturation
i5
pressure
i6
fluid (polymer) extra stress
i7
voltage
A value of 0 for an integer directs Goma to exclude contributions from that variable in the error norm calculation; correspondingly, a value of 1 means that variable should be included.
Examples#
A sample time step error card follows:
Time step error = 0.01 0 1 1 1 0 0 0
In this example, the L2 norms for the fluid velocity, temperature, and concentration are summed (and scaled) prior to comparison with the target error value of 0.01. If the norms of the velocity, temperature, and concentration variables is greater than 0.01, the time step is halved and the step repeated. Otherwise, the current step size is compared to other step criteria before continuing to the next step.
If the integer values are omitted, the scaled error norm becomes infinite and the analysis will terminate in the error norm calculation with an arithmetic overflow.
Examples#
To use the normalized value of the norm, the following would be specified:
Time step error = -0.01 0 1 1 1 0 0 0
This would set the maximum time step error to be 1%.
Technical Discussion#
Note that on porous flow problems the error in step-size is computed as a composite measure of all porous-flow variables, viz. these cannot currently be controlled separately.